[EM] Methods, and the criteria they satisfy

Diana Galletly dag1000 at eng.cam.ac.uk
Thu Sep 18 23:56:02 PDT 2003


[Apologies if this appears more than once; I tried to send it over
12 hours ago and it didn't appear (and isn't in the archive either)
so I'm guessing it didn't get posted; I'm resending because I really
could do with an answer to my questions!]

I'm trying to write an article about suitable voting systems for
referenda (one proposal, which then has amendments suggested, and
the vote is currently conducted using IRV).  Recently there have
been some potentially dubious results, and I've been looking into
other methods which might be preferable.  What I'd like to do is
draw up a table showing which systems satisfy which criteria.
I've read Mike Ossipoff's, Blake Cretney's and Stephen Eppley's
websites, and find myself somewhat confused as to which criteria
are the same between the webpages, and which are different.

The methods I want to compare are Approval, Average Rating, Borda,
IRV, Plurality and Condorcet (with assorted completion
rules: CpSSD, MAM, and Tideman's Ranked Pairs at the very least,
and possibly a scheme of my own devising -- of which possibly
more later).  The criteria I wish to compare these systems against
are the Condorcet Criterion, the Independence of Clones Criterion (ICC)
the Local Independence from Irrelevant Alternatives Criterion (LIIAC), the
monotonicity criterion, the Smith criterion, Cretney's Secret
Preferences Criterion (SPC) and Consistency Criterion, Ossipoff's
Generalised Strategy-Free Criterion (GSFC), Strong Defensive Strategy
Criterion (SDSC) and Weak Defensive Strategy Criterion (WDSC).  Some
of these have different names on Eppley's pages.

I'd appreciate corrections on the following beliefs:

1). Approval and Average Rating meet ICC, Consistency Criterion, monotonicity
and WDSC, fail SDSC and GSFC, and the other criteria are either irrelevant or
failed.
2). Borda meets only Consistency Criterion and monotonicity and fails the
rest.
3). IRV meets only ICC and SPC.
4). Plurality meets consistency and SPC (although that could be argued to be
irrelevant); the other criteria are either irrelevant or failed.
5). All the Condorcet methods meet the Condorcet Criterion, ICC, LIIAC,
monotonicity, the Smith Criterion and GSFC, and fail the Consistency
Criterion and SPC.  I've read that Tideman's Ranked Pairs fails SDSC and
WDSC, but the other two completion methods meet them.

Am some (or all) of these beliefs incorrect?  I'd be very grateful for
any help that people could give me.

Diana.



More information about the Election-Methods mailing list